Pre-event reputational crisis: how it's spotted before it breaks.

It happened this week: a Chief of Comms at a regional client sent us a message at 11 at night. Their brand was on the front page of a national newspaper. A reputational crisis, two ministers weighing in, a hashtag climbing. The question wasn't how to put out the fire. The question was how nobody had seen it coming.

The thing is, they had seen it. Three months earlier, the team had detected three NGOs coordinating messages about the brand. A policy paper from a LATAM think tank, two months back. Two trade outlets, three weeks ago. It was all there. What was missing was the read that connects those four dots into a line.

When a reputational crisis reaches the national newspaper, it's already a crisis. There's no room, no preparation, no narrative control. All that's left is containment. The hard part — and the valuable part — is reading it while it's still a weak signal. Three accounts, an NGO cluster, a think-tank paper, a creator with a small audience building a case.

Case 01 · The signal that three months later became a headline

FMCG beverages sector. Brazil. Retroactive detection, May 2025.

Three environmental NGOs began coordinating messages about a component of the brand's packaging. 47 mentions in the first week, all with a small audience — between 800 and 4,000 followers per account. For a global dashboard, it was noise. Low volume, no media, no major influencers, negative sentiment but within the usual range.

What the dashboard didn't see: the three NGOs had coordinated two prior campaigns together, both ending in regulatory pressure. One of them had an internal brief leaked six months earlier naming the category as the year's target. And the 47 messages — read one by one — used the same framing, the same three keywords, almost the same argumentative order.

Three weeks later, a LATAM think tank published a policy paper on the component. 200 reads. Two trade outlets picked it up. By the time the national newspaper covered it, the cluster already had 14 verified accounts aligned, two members of congress retweeting, a consumer association drafting an open letter. Three months between the first signal and the headline. The brand had 90 days to prepare a response. It didn't use them.

Case 02 · The cluster that needed connecting

Banking sector. Argentina. October 2025.

The client called us because a creator with 80,000 followers had put together a viral thread on "the end of the honeymoon with digital banking". 4 million impressions in 48 hours. The CMO's question: where did this come from?

It came from three different places the internal team had open in separate tabs. Individual complaints about fees since April. Complaints about the app since July. Complaints about branch service since August. For the dashboard, three separate topics. For the senior read, all three came from the same segment: young people in Buenos Aires and Greater Buenos Aires, digital-banking-first, aspiring to their first mortgage.

When a creator articulates a narrative that was already latent, they don't create it: they name it. They give it a title. And from that title on, the rest of the actors who were watching the same clusters separately now have a common name to connect them. That's when the conversation scales. Not before.

The detail is that this convergence was in the data since July. Three clusters, same segment, same tone. What was missing was the topic-and-demographics cross-reference sustained over time. A dashboard won't do it on its own because it wasn't designed for that. An analyst who knows the business will.

Case 03 · The paper nobody read

Pharma sector. Mexico. January 2026.

An independent consultant published a policy paper on the prices of imported medicines. A 24-page PDF, 200 reads in the first month. Three brands named by name in the body of the paper. All three had active monitoring. None of them caught it.

Try it · would you read it as signal or noise?

Three real posts. You decide the verdict.

Mariana O. @mariana_pol · Mexico City · 6h

𝕏

just read @consultor_mx's paper on the prices of imported medicines. 24 pages, devastating. names 3 pharma companies. this is going to be a topic in congress.

💬 8 🔁 14 ♥ 32 📊 1.840

Noise Weak signal Full-blown crisis

Weak signal. Mariana is a legislative advisor with 4K followers. Low volume, high capacity for institutional traction. When she anticipates "this is going to congress", it usually goes to congress. The dashboard counts it as one mention. The senior read flags her as a propagation vector.

Ricardo F. @ricky_gdl · Guadalajara · 2d

𝕏

terrible service at the pharmacy today, an hour waiting for a medicine they didn't even have. and then they say prices are high for a reason.

💬 2 🔁 0 ♥ 4 📊 187

Noise Weak signal Full-blown crisis

Noise. An individual complaint, no coordination, no actor with traction, a reactive emotional register. A thousand of these a day in the category. Worth adding to the sentiment, but not to the alert. Here the global dashboard gets it right.

Chronic Patients Assoc. MX @asocpac_mx · National · 3d

𝕏

we're revisiting the figures from @consultor_mx's paper on import prices. three labs account for 62% of the annual adjustments above CPI. we keep requesting a hearing with Cofepris.

💬 19 🔁 47 ♥ 118 📊 6.420

Noise Weak signal Full-blown crisis

Weak signal with a cluster forming. An organized association citing a technical paper, requesting a regulatory hearing, naming concentration with specific numbers. Low volume but institutional structure. It's the second actor in the paper's cycle. The third will be a legislator.

Three posts. Two are signal. One is noise. The hard part isn't detecting them — it's deciding which one makes it to Monday's table.

Three months after the paper was published, the Chronic Patients Association cited it in its request for a hearing before Cofepris. A month later, a legislator picked it up in a committee. Another month, a story in a national outlet. The brands named in the paper received a written letter six months after publication. If someone had read the paper in January — someone with the judgment to understand that a consultant with a regulatory track record publishing about your category isn't noise — today those brands would be the brands with a ready response. They weren't.

What is a weak signal?

After three years doing this across regional brands, we arrived at three operational properties. It's not an academic definition, it's a working filter. If all three are present, it isn't noise.

The three properties:

1. Low volume. Dozens, not thousands. This is where most systems don't look — because they're calibrated to alert on spikes, not on patterns. But the spike comes later. The shape comes first.

2. Specific audience. Not general. If everyone is saying it, it's no longer a weak signal — it's something else. A weak signal has identifiable authorship: NGOs, think tanks, associations, consultants with a regulatory track record, creators with a small audience but aligned with a specific public.

3. Convergence. Several not-yet-connected actors talking about the same topic. When they converge, it's no longer a signal: it's a headline. The window between the two is where the read is won or lost.

What a Chief of Comms told me a few months ago, after having to react to a crisis the team had seen coming three times before:

By the time it reached the table, it was already a crisis. The signals had been there for 90 days. The problem wasn't that we didn't have them — it was that nobody read them for what they were.

That's the line. It's not a data problem — there's plenty of data. It's a reading problem: someone with business judgment who can look at 47 mentions and say "this, in 90 days, is going to be the front page". And someone else — or the same person, if they have the authority — who can take that read to leadership and get them to act on it while there's still room.

Here's what happens: most listening platforms are built to alert when something has already happened. Spikes, anomalies, sharp shifts in sentiment. They're useful, but they arrive late to the pre-event reputational problem. What's missing — and what we do when we work with a regional client — is reading the conversation before the spike, with judgment about who's talking, with what track record, connected to which other actor, within what time window.

That read isn't delivered by a dashboard. It's delivered by people who know the category, the geography and the actors. And the output isn't an alert — it's a decision: we act now or we wait. When it's done well, the brands we work with don't end up in headlines they didn't choose. When it's not, they end up there.